measurement process capability - part 1 · measurement process capability - part 1 you are scarcely...
Post on 15-Jul-2020
17 Views
Preview:
TRANSCRIPT
1FIRST PUBLISHED IN AUGUST 2017
PIQ-ONLINE
There is still need for discussion – a working paper
MEASUREMENT PROCESS CAPABILITY - PART 1STEPHAN CONRAD | Q-DAS GMBH
Nowadays, measurement process capability according to VDA Volume 5 and/or ISO 22514-7 is well-established. The Volkswagen group (VW, Audi, Seat, Skoda…) adapted their VW 10119 guideline years ago, the LF5 Daimler guideline is already based on the latest edition of VDA Volume 5, BMW modified the group standard 98000 accordingly, Bosch updated booklet 8… however, this fact alone does not answer all the questions. This article discusses some current aspects that leave room for interpretation since they are not based on any “official” regulations. Read part 1 of this series of articles.
MEASURING SYSTEM N.O.K., MEASUREMENT PROCESS O.K.?
Page 42 of VDA Volume 5 says: “ The limits for the capability
of measuring systems and measurements processes must
be determined. It is important to consider that the influences
of the form deviation of test parts can affect the evaluation of
the measurement process considerably. It is recommended
that the capability ratio for measuring systems, QMS_max
amounts to 15% and, for measurement processes, QMP_max
amounts to 30%.”
2 FIRST PUBLISHED IN AUGUST 2017
MEASUREMENT PROCESS CAPABLE BUT STILL USELESS?
Even the limit of QMP ≤ 30% has already caused some
debates. There is no doubt that measurement engineers
are hardly able to observe it. For now, let us look at the
consequences this limit has.
QMP indicates the uncertainty variation range based on
a “probability of 95%”. Strictly speaking, the probability
amounts to 95.45%; broadly speaking it is also referred to
as “4s”. Loosely speaking, this is the “variation” caused by the
measurement process alone. It is a kind of noise floor caused
by the measurement process alone. Let us assume you start
a machine acceptance test and expect Cm/Cmk ≥ 2.0. Even if
the production facility does not cause any variation at all and
you identify nothing but the “noise floor” of the measurement
process, you will have to include the 99.73% variation range
(“6s”) of the measurement uncertainty in your calculation of
Cm. Believe it or not, in case of QMP = 30%, this range amounts
to 45% of the tolerance and thus leads to Cm = 2.22. The
scope left for the machine is limited. The analysed machine
needs to reach a capability of Cm ≥ 4.6 under real conditions
in order that the evaluation sheet shows Cm ≥ 2 in the end.
This is just utopian. And what is the consequence? A capable
measurement process does not have to be capable for a
machine capability analysis.
SPECIFY AN APPLICATION-SPECIFIC LIMIT
How shall we deal with this aspect? It is not enough to simply
lower the limit since the causes of this problem are complex.
The limits VDA Volume 5 recommends are interpreted in a
way that the capability ratio for the measuring system shall
be QMS ≤ 15% while the capability ratio of the measurement
process shall amount to QMP ≤ 30%. VDA Volume 5’s
flow chart (Figure 7, p. 41) supports this point of view
assuming that half the uncertainty budget is caused by the
measuring system and half of it comes from other influence
components of the measurement process. (Whoever wants to
assess specific figures has to consider that the calculation
of measurement uncertainties is based on a quadratic
addition.)
Nonetheless, the measuring system frequently exceeds its
budget (e.g. QMS = 17% > 15%) while the total process does
not even reach its limit (e.g. QMP = 24% < 30%) and is thus
capable. Especially characteristics with tight tolerances and
standards/reference parts having a relatively high calibration
uncertainty that is physically hardly feasible typically cause
problems.
The question is whether the measurement process as a
whole has proved to be not capable, and justifiably so,
since it is not important at all, at least in most use cases,
whether influences from the measuring system or from
the measurement process dominate the measurement
uncertainty U = 12%.
A possible solution is to leave out the evaluation of QMS. You
only use QMP to evaluate an existing measurement process.
Another possibility is to raise the limit QMS in general. This
might, however, cause inconsistencies, especially when
you evaluate a measuring system to make a preselection.
The limit of QMS ≤ 15% will then be fairly appropriate to
have reasonable scope left for the hitherto unknown
measurement process.
3FIRST PUBLISHED IN AUGUST 2017
MEASUREMENT PROCESS CAPABILITY - PART 1
you are scarcely able to respect the limit, any subsequent
acceptance tests for the measurement process affected
by the operator and series production parts will be likely
to exceed this limit on site. It might thus be reasonable to
assign a suitable limit to the measurement process model.
A flexible model might even make it easier to align typical
MSA methods with VDA Volume 5. In historical terms,
standards and guidelines usually demanded
• a type-1 study based on Cg /Cgk ≥ 1.33 and
• a type-2 study based on %GRR ≤ 10%,
however, we could now interpret this requirement as
“MSA(V1/V2)” measurement process model based on special
limits.
• Applying usual boundary conditions, i.e.
• RE ≤ 5%T
• UCAL ≤ 5%T
• BI ≤ 5%T
• 99.73% MSA variation range,
we will expect QMS ≤ 19% and QMP ≤ 20% .
• Applying e.g. Bosch booklet 10 (2015) and expecting
calibrated reference parts having a low measurement
uncertainty and a bias that is as small as possible, i.e.
• RE ≤ 5%T
• UCAL ≤ 1%T
• BI = 0
• 99.73% MSA variation range
the limits shall be adjusted to QMS ≤ 12% and QMP ≤ 13%.
VDA Volume 5 wants to determine the uncertainty of “the
measurement process” and thus evaluates all influence
quantities affecting the measurement process. The question
is whether the same influence quantities affecting a short-
term machine capability analysis also affect a long-term
analysis in terms of statistical process control. If this is not
the case, the performance of a measurement process in a
machine capability analysis will not be the same as for the
purpose of SPC, even though the technical equipment and
the characteristic of the part to be measured are identical.
VDA Volume 5 takes this aspect into account by giving an
overview of typical measurement process models and
associated typical uncertainty components in chapter 7.1.
We do not know how closely the models given in VDA Volume
5 reflect real requirements, but the concept behind them
is brilliant. The only crux is the standardised limit. As an
example, when you agree on an acceptance test according to
model C with the manufacturer of the measuring system and
MEASUREMENT PROCESS CAPABILITY - PART 1
• Focusing on Mercedes Benz Cars LF5 (2010) specifying
• RE ≤ 5%T
• UCAL ≤ 5%T
• BI = 0
• 99.73% MSA variation range
the limits are QMS ≤ 18% and QMP ≤ 22%.
These simple examples show that the system is highly
flexible.
Q-DAS GmbH
Eisleber Str. 2
69469 Weinheim
www.q-das.de
stephan.conrad@hexagon.com
Interested in this topic?
top related